A NOTE ON THE PROOF OF
BERTRAND'S THEOREM
Vladimir Jovanovi¢
Abstrat. In this paper well a ommon gap inthe proof of Bertrand'
theorem present both the inBertrand's originalpaper Théorème relatif au
movementd'unpointattiréversunentrexeandintheArnold'sbook
Math-ematialmethodsof lassial mehanis, byprovidingmissing detailswhih
pertaintotheproblemofhowtosingleoutelastiandgravitationalpotentials
amongthepowerlawones.
1. Introdution
Motion of a partile with a unit mass in a entral eld is desribed by the
Newton's equations
(1.1)
¨
−
→
r
=
−
∂U
∂
−
→
r
,
where
−
→
r
=
−
→
r
(t)
∈
R
3
istheradiusvetorofthepartileat time
t
,andU
=
U
(r)
is the potential energy or simply the potential; herer
=
|−
→
r
|
is the Eulidean
norm of the vetor
−
→
r
. It is a ommon fat that everymotion in a entral eld
atually ours in a xed plane. If we denote by
π
the plane in whih a given motion ours, then we an write−
→
r
=
r(cos
ϕ
−
→
e
1
+ sin
ϕ
−
→
e
2
)
, where
−
→
e
1
,
−
→
e
2
is anorthonormal basis of
π
, and(r, ϕ)
are polar oordinates in this plane. Let us denote−
→
e
r
=
−
→
r
r
= cos
ϕ
−
→
e
1
+ sin
ϕ
e
→
−
2
. Then (1.1)anberewrittenas (1.2)¨
−
→
r
=
−
U
′
(r)
−
→
e
r
.
If
−
→
e
ϕ
=
−
sin
ϕ
−
→
e
1
+ cos
ϕ
−
→
e
2
, then˙
−
→
e
r
= ˙
ϕ
−
e
→
ϕ
and
˙
−
→
e
ϕ
=
−
ϕ
˙
−
→
e
r
. Therefore,−
→
r
=
r
−
→
e
r
and(1.2)imply
(1.3)
r
¨
−
r
ϕ
˙
2
=
−
U
′
(r),
2 ˙
r
ϕ
˙
+
r
ϕ
¨
= 0.
Fromtheseond equationin(1.3)follows
r
2
ϕ
˙
=
M,
2010MathematisSubjetClassiation: 70B05,34A34.
foraertainonstant
M
. Therstequationin (1.3)wetransforminto(1.4)
r
¨
=
−
V
′
(r),
where
V
(r) =
U
(r) +
M
2
2r
2
is the amended potential. From (1.4) follows that theenergyofthesystem(1.1)isaonstantofthemotion:
1
2
r
˙
2
+
V
(r) =
E.
Ouronsiderationswillhavemuhtodowithirularorbits: thesearesolutionsof
(1.1) with
r
=
ρ
onstant. From(1.4)wededuethatV
′
(ρ) = 0
isneessaryand suient ondition forr
=
ρ
to bea irular orbit. Assume thatV
′′
(ρ)
>
0
. In [3, subsetions12.6and12.7℄is developedatheorywhihassertsthat foragivenenergy level
E > V
(ρ)
lose toV
(ρ)
, the solutions of (1.4) periodially osillate in theinterval[r
min
, r
max
]
∋
ρ
(see Figure 1). The position of the partile, whenr
=
r
min
isalledaperienter andforr
=
r
max
anapoenter. TheangleΦ
between twoonseutiveperientersandapoenters,thesoalledapsidalangle,isgivenby(see[2℄)
(1.5)
Φ =
Z
r
max
r
min
M/r
2
p
2(E
−
V
(r))
dr.
A simplebutfundamental observationisthatanorbitof (1.1)islosedifand
onlyif
Φ/π
∈
Q
.r
r
min
r
max
E
V
Figure 1
Already Newton knew that all bounded orbits for elasti and gravitational
potentials are ellipses (see [9℄). However, if there are other entral potentials
whih give rise to losed orbitsunder arbitrary initial onditions(provided they
are bounded) hadbeenunknown for almost twoenturiesuntil Bertrand in 1873
published hispaper[4℄inwhihheprovedthat theanswerwasno. Thisassertion
istodayknownasBertrand's theorem.
A lotofeorthasbeenspenttoextendBertrand'stheorem tothegeometries
and spaesotherthantheEulideanthreedimensionalone: forexample,in [8℄it
isprovedfortheLobahevskyandspherialgeometry,andin[5℄,[10℄ forsurfaes
ofrevolution. LetusstressthatallthesegeneralizationsarebaseduponBertrand's
originalproof. Forfurtherdevelopmentin thisdiretionsee[11℄.
Some authors used quite dierent tools to handle the problem, in partiular
perturbation theory of integrableHamiltonian systems. For example,Albouy [1℄
and Féjoz,Kazmarek[6℄ appliedBirkhonormalforms alulatedalongirular
Bertrand's original proof anbedivided into three steps: rst, he alulated
thelimitoftheapsidalanglewhentheorbitapproahestheirularone,andthen
employingontinuousdependene oftheapsidalangleonorbitsheestablishedits
independene onenergy,whihonsequentlyimpliesthattheangleisequaltothe
aforementionedlimit. In the seond step is used the fat that the apsidal angle
is independent on momentum aswell, and, as a onsequene, one arrivesto the
powerlaw forthe potential:
U
(r) =
ar
α
. Thethird stepdealswith the problem
of howto singleoutthegravitationalandelasti potentialsamongthe powerlaw
ones. Bertrand resolvedthis stepessentiallyby takingsome speial limits in the
initial form ofthe apsidalangle. This stepofthe Bertrand'sproofis alsotreated
inArnold'sbook[2℄. However,inbothasesdetailslakanditturnsoutthatthey
arenotsoeasytoprove. Thisisexatlythepurposeofthepaper: toprovidesome
details missing in[4℄ and[2℄in theproofofthis step. Itis theontentofLemma
(2.1).
An originalapproahto thethird stepbasedonthefrationalalulusanbe
foundin [7℄.
2. Formulation and proofof Lemma
Lemma 2.1. If
U
(r) =
ar
α
and
a >
0
,thenlim
E
→∞
Φ =
π
2
,
for all
α >
0
.Proof. Letusintrodueanewvariable
x
byM
r
=
x.
Dene thefuntion
W
byW
(x) =
V
(r)
,thatisW
(x) =
1
2
x
2
+
U
M
x
=
1
2
x
2
+
1
2
kx
−
α
,
where
k
= 2aM
α
. Let
x
1
=
M/r
max
andx
2
=
M/r
min
. Following[2℄weapplythe substitutionx
=
yx
2
in (1.5)andgetΦ =
Z
1
y
1
dy
p
H
(y)
,
where
y
1
=
x
1
/x
2
andH(y) = 1
−
y
2
−
kx
−
α
−
2
2
(y
−
α
−
1).
Notethat
H
′
(y) = 0
onlyfory
=
y
∗
,where(2.1)
y
∗
=
lx
−
1
2
,
with
l
= (αk/2)
1/(α+2)
and
H
′′
y
0
y
1
y
∗
1
H
Figure 2
Consequently,
H(y)
>
H
(y
∗
)
−
H
(y
1
)
y
∗
−
y
1
(y
−
y
1
) +
H
(y
1
),
y
∈
(y
1
, y
∗
),
H
(y)
>
H
(y
∗
)
−
H
(1)
y
∗
−
1
(y
−
1) +
H
(1),
y
∈
(y
∗
,
1),
(seeFigure 2)and
H
(y
1
) =
H
(1) = 0
imply(2.2)
1
p
H
(y)
<
r
y
∗
−
y
1
H
(y
∗
)
1
√
y
−
y
1
,
y
∈
(y
1
, y
∗
),
and
(2.3)
1
p
H
(y)
<
s
1
−
y
∗
H
(y
∗
)
1
√
1
−
y
,
y
∈
(y
∗
,
1).
It isalso
(2.4)
1
p
1
−
y
2
<
1
p
H(y)
,
y
∈
(y
1
,
1).
Due to
r
min
→
0
,r
max
→ ∞
asE
→ ∞
,itisx
1
→
0+, x
2
→ ∞
,whene(2.5)
y
1
→
0, y
∗
→
0
H
(y
∗
) = 1
−
y
2
∗
−
kl
−
α
x
−
2
2
+
kx
−
α
−
2
2
→
1.
TheproofoftheLemmaisbaseduponthefollowinginequality
(2.6)
Z
1
y
1
dy
p
H
(y)
−
Z
1
y
1
dy
p
1
−
y
2
6
Z
1
y
∗
dy
p
H(y)
−
Z
1
y
∗
dy
p
1
−
y
2
+
Z
y
∗
y
1
dy
p
H(y)
+
Z
y
∗
y
1
dy
p
Evidently,
R
y
1
dy
√
1
−
y
2
→
π
2
andR
∗
y
1
dy
√
1
−
y
2
→
0
. The inequalities (2.4) and (2.3),implyZ
1
y
∗
dy
p
H
(y)
−
Z
1
y
∗
dy
p
1
−
y
2
=
Z
1
y
∗
p
1
−
y
2
−
p
H
(y)
p
H(y)
p
1
−
y
2
dy
=
Z
1
y
∗
1
−
y
2
−
H
(y)
p
H(y)
p
1
−
y
2
(
p
H(y) +
p
1
−
y
2
)
dy
=
kx
−
α
−
2
2
Z
1
y
∗
y
−
α
−
1
p
H
(y)
p
1
−
y
2
(
p
H
(y) +
p
1
−
y
2
)
dy
6
kx
−
α
−
2
2
Z
1
y
∗
y
−
α
−
1
H
(y)
3/2
dy
6
1
−
y
∗
H(y
∗
)
3/2
kx
−
α
−
2
2
Z
1
y
∗
y
−
α
−
1
1
−
y
1
√
1
−
y
dy.
Thelasttermtendsto
0
: indeed, owingto(2.5),itis1
−
y
∗
H(y
∗
)
3/2
→
1,
and thanks to
y
−
α
−
1
< y
−
α
6
y
−
α
∗
fory
>
y
∗
, and the equalityx
2
y
∗
=
l
(see (2.1)),respetively,itfollowskx
−
α
−
2
2
Z
1
y
∗
y
−
α
−
1
1
−
y
1
√
1
−
y
dy
=
kx
−
α
−
2
2
Z
1
1/2
y
−
α
−
1
1
−
y
1
√
1
−
y
dy
+
kx
−
α
−
2
2
Z
1/2
y
∗
y
−
α
−
1
1
−
y
1
√
1
−
y
dy
6
kx
−
α
−
2
2
Z
1
1/2
y
−
α
−
1
1
−
y
1
√
1
−
y
dy
+
kx
−
α
−
2
2
Z
1/2
y
∗
y
−
α
∗
(1
−
y
2
)
3/2
dy
=
kx
−
α
−
2
2
Z
1
1/2
y
−
α
−
1
1
−
y
1
√
1
−
y
dy
+
kl
−
α
x
−
2
2
Z
1/2
y
∗
dy
(1
−
y
2
)
3/2
→
0.
Thus, the rst expression on the r.h.s. in (2.6) tends to
0
. It only remains to provethesamefor theseond termonr.h.s.of (2.6). Itis, however,asimpleonsequeneof (2.2):
Z
y
∗
y
1
dy
p
H(y)
6
r
y
∗
−
y
1
H(y
∗
)
Z
y
∗
y
1
dy
√
y
−
y
1
→
0.
Aknowledgments. I want to thank prof. Alain Albouy for his beneial
Referenes
1. A.Albouy,Leturesonthetwo-bodyproblem,ClassialandCelestialMehanis: TheReife
Letures,PrinetonUniversityPress,Prineton,2002,63116.
2. V.I.Arnold,Mathematialmethodsoflassialmehanis,Springer-Verlag,NewYork,1989.
3. ,Ordinarydierentialequations,MITPress,1978.
4. J.Bertrand, Théorème relatifau movementd'un point attiréversunentrexe,Comptes
rendus77(1873),849853.
5. G. Darboux, Étude d'une question relative au mouvement d'un point sur une surfae de
révolution,BulletindelaS.M.F.5(1877),100113.
6. J.Féjoz,L.Kazmarek,SurlethéorèmedeBertrand(d'aprèsMihaelHerman),Ergod.Th.
&Dynam.Sys.24(2004),17.
7. Y.Grandati,A.Bérard,F.Ménas,InverseproblemandBertrand'stheorem,Am.J.Phys.76
(2008),782787.
8. H.Liebmann,ÜberdieZentralbewegunginderniheuklidishenGeometrie,LeipzigBerihte
55(1903),146153.
9. I.Newton,Philosophiaenaturalisprinipiamathematia,London,1687.
10. V.Perlik,Bertrandspaetimes,Class.QuantumGrav.9(1992),10091021.
11.
O. A. Zagrdski, E. A. Kudrvceva, D. A. Fedoseev
,Obobwenie teorem y Bertrana
O DOKAZU BERTRANOVE TEOREME
Rezime.
U ovom radu dopunjujemo dokaze Bertranove teoreme iz
Bertrano-vog originalnog rada
Théorème relatif au movement d'unpoint attiré vers unentre xe
, i Arnoljdove knjige Matematiqke metode klasiqne mehanike,
predstavljajui nedostajue korake koji omoguuju da se elastiqni i
gravitacioni potencijal izdvoje meu potencijalima oblika stepene
fu-nkcije.
FaultyofSienesandMathematis (Reeived27.04.2015)
UniversityofBanjaLuka (Revised26.06.2015)
BanjaLuka
RepubliofSrpska,BosniaandHerzegovina